mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jie Yu <yujie....@gmail.com>
Subject Re: Review Request 43015: Fixed persistent volumes with docker tasks.
Date Tue, 16 Feb 2016 23:46:06 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43015/#review119340
-----------------------------------------------------------




src/slave/containerizer/docker.cpp (lines 161 - 162)
<https://reviews.apache.org/r/43015/#comment180653>

    Can you wrap using 70 char width here?



src/slave/containerizer/docker.cpp (lines 413 - 414)
<https://reviews.apache.org/r/43015/#comment180687>

    Should we ignore those volumes that have absolute/nested container path (like we did in
linux fs isolator). What's the reason removing it?



src/slave/containerizer/docker.cpp (line 419)
<https://reviews.apache.org/r/43015/#comment180662>

    Indentation.



src/slave/containerizer/docker.cpp (line 442)
<https://reviews.apache.org/r/43015/#comment180688>

    indentation



src/slave/containerizer/docker.cpp (lines 457 - 458)
<https://reviews.apache.org/r/43015/#comment180690>

    Ditto. Why remove the container_path check?



src/slave/containerizer/docker.cpp (lines 473 - 488)
<https://reviews.apache.org/r/43015/#comment180699>

    There's a os::exists check in linux fs isolator which you removed. This is fine because
we won't update volumes when containerizer->update is called. I think we should at least
drop a TODO here
    so that we don't forget that when we want to support containerizer->update.
    
    Alternatively, we can kill updatePersistentVolumes and inline it into mountPersistentVolumes.



src/slave/containerizer/docker.cpp (line 881)
<https://reviews.apache.org/r/43015/#comment180700>

    Can you add a TODO stating that this is a hack for now. We need to revisit if docker containers
have other mounts that may container 'containerId'



src/slave/containerizer/docker.cpp (lines 918 - 921)
<https://reviews.apache.org/r/43015/#comment180703>

    Indentation.
    ```
    futures.push_back(
        docker->stop(
            container.id,
            flags.docker_stop_timeout,
            true)
          .then([id]() { return id.get(); }));
    ```



src/slave/containerizer/docker.cpp (lines 1701 - 1708)
<https://reviews.apache.org/r/43015/#comment180694>

    Why do you need to do this? Looks like we don't mount until pulling is done. Am I missing
something?



src/slave/containerizer/docker.cpp (lines 1785 - 1790)
<https://reviews.apache.org/r/43015/#comment180704>

    Hum, I think we shouldn't remove the volumes if the container might be still running.
So can we move the cleanup logics to `___destroy()`?



src/tests/containerizer/docker_containerizer_tests.cpp (lines 1178 - 1217)
<https://reviews.apache.org/r/43015/#comment180721>

    ok, can you instead, launch a slave to do that. This is pretty hacky to me.
    
    You can split this test into two: 1) for running containers 2) for orphan containers.
    
    Take a look at SlaveRecoveryTest see how to simulate slave failover.



src/tests/containerizer/docker_containerizer_tests.cpp (lines 3217 - 3221)
<https://reviews.apache.org/r/43015/#comment180708>

    You can do CREATE+LAUNCH in the same offer cycle, right?



src/tests/containerizer/docker_containerizer_tests.cpp (line 3236)
<https://reviews.apache.org/r/43015/#comment180713>

    I might want to write some data to the persistent volume and make sure that file exists
in the end. See AccessPersistentVolume in PersistentVolumeTest.



src/tests/containerizer/docker_containerizer_tests.cpp (line 3268)
<https://reviews.apache.org/r/43015/#comment180715>

    insert a blank line above.



src/tests/containerizer/docker_containerizer_tests.cpp (line 3270)
<https://reviews.apache.org/r/43015/#comment180714>

    insert a blank line above.



src/tests/containerizer/docker_containerizer_tests.cpp (line 3285)
<https://reviews.apache.org/r/43015/#comment180712>

    Does this still work? Looks like you're not doing rmdir anymore. Am i missing something?


- Jie Yu


On Feb. 16, 2016, 3:33 a.m., Timothy Chen wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/43015/
> -----------------------------------------------------------
> 
> (Updated Feb. 16, 2016, 3:33 a.m.)
> 
> 
> Review request for mesos and Jie Yu.
> 
> 
> Bugs: MESOS-3413
>     https://issues.apache.org/jira/browse/MESOS-3413
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Fixed persistent volumes with docker tasks.
> 
> 
> Diffs
> -----
> 
>   src/slave/containerizer/docker.hpp 77a50d80179672cf3c270cbdd7fa003c7d9ad6f3 
>   src/slave/containerizer/docker.cpp ed1c9a551f03a37d572470e4c495f5df834198cc 
>   src/tests/containerizer/docker_containerizer_tests.cpp 645bdcf095145097d8b8c65d592c787417883145

> 
> Diff: https://reviews.apache.org/r/43015/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Timothy Chen
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message